A fast direct Fourier-based algorithm for subpixel registration of images

Abstract

Abstract-This paper presents a new direct Fourier-based algorithm for performing image-to-image registration to subpixel accuracy, where the image differences are restricted to translations and uniform changes of illumination. The algorithm detects the Fourier components that have become unreliable estimators of shift due to aliasing, and removes them from the shift-estimate computation. In the presence of aliasing, the average precision of the registration is a few hundredths of a pixel.Experimental data presented here show that the new algorithm yields superior registration precision in the presence of aliasing when compared to several earlier methods and has comparable precision to the iterative method of Thévenaz et al. [21].

Optimal Registration Of Aliased Images Using Variable Projection With Applications To Super-Resolution

“…This is perhaps the most commonly utilized motion model for aliased image registration algorithms [16]. Using this assumption, we can simplify the problem and produce a very fast algorithm.…”

“…We applied the proposed multiframe registration algorithm using Equation (6) as the cost function. We compared this algorithm with the relative phase approach [16], a standard iterative gradient-based approach for estimating global translation [15] and a cyclic coordinate descent type algorithm as described earlier similar in spirit to [11,24]. The tuning parameters were modified to provide the best MSE possible at each image scale.…”

“…Relative phase [16] Algorithm PSD using a periodogram method [32] applied to the high resolution large image shown in Fig. 3.…”

“…7 shows the reconstructed images using both the relative phase [16] and gradient-based [15] pairwise registration algorithms. While the gradient-based approach is clearly superior to the relative phase approach, both reconstructions exhibit slightly jagged edges due to inconsistent motion estimates.…”

“…This assumption obviously breaks down in the presence of aliasing. Again, the robust phase estimation algorithms of [16,17] make assumptions about the spectral decay of a typical image. The effects of such bias in pairwise algorithms have been noticed motivating robust approaches to minimize the effects of poor registration estimates on the final image reconstruction [13].…”